Lost so many matches

[1_mke]

The thing is, without secondary data like cross referencing matches I don't think there is any way to be sure whether matches below a certain level of "quality" are real or not and the matching process already takes a lot of resources to complete.

Given that FTDNA has to draw the line some place where it feels confident that most of the matches on the wrong side of the line are noise. Yes, they are going to end up throwing out some good matches. Without investing a lot more programming and computational time I don't see how that can be minimized further.

[djknox]

Perhaps said in jest, but more valuable than you realize Kasandra! At least in semi-rural places in Ireland. And the phone book can be contemporary and online. On a few occasions I have written into what I call "the void" based on addresses I have found in the Irish phone book and have gotten responses from near-to-distant relatives and even some incredibly useful information from those who aren't. In my mailings I include organized charts, civil registration and census records, etc, and an email for response to make it easy.

Hey I would love to find online phonebooks of individual Irish villages! Can you suggest where to find these online??

[jpulla]

I lost also many matches, but extremely disappointing was losing my 5th cousin who was confirmed by paper trail. All 4 over 5cM segments have gone, he is not any more my cousin, not even speculative, in my FF-matches. The biggest segment with him was before the build37 upgrade over 9 cM. 23andme shows still all 4 segments with him. Sounds weird. How I can know how many cousins being in range of 4th-distant cousins have not been reported?

[blejerh]

Among the matches lost were my 3rd1R cousin and my 5th1R cousin. The 3rd cousin matched my son but not me. The 5th1R cousin matched both of us.

As I have mentioned before, I have a lot of DNA matches that I share with my son which show up on Gedmatch but not on FTDNA, due to the threshhold or algorithm. Since my son is 75% Ashkenazi Jewish and I am 50%, his overall shared DNA is larger, even though the segments we share are either identical or mine are longer.

I suspect that these vagaries of the algorithm and threshhold will eventually be better understood and "fixed." In the meantime, Gedmatch is my friend because I am unable to use FTDNA to determine that someone who matches my son but not me actually is my husband's ancestor. In most of the instances that I concluded that, I found it was incorrect once I went over to Gedmatch.

[blejerh]

Sorry, should have said "shares ancestry with my husband"...

[sjadelson]

On the other hand, many of my spurious matches (matched more strongly to me than to my father) have been downgraded such that they show much more distantly to me. Overall, a win.

[Carolus]

23andMe updated Build 37 last year and I could not see any dramatic changes.Which one is now more reliable? I think that the both labs have different systems and wondering how does this affect to comparisions on Gedmatch?

[jbarry6899]

Pm

Djknox--sent you a private message on Irish phone books.

Jim

[ursus]

I thought that the following might be interesting. I am an Ashkenazic Jew who had 1711 matches before the update and gained 124 new ones but lost many others, so I am down to 1425. (Conceivably there are matches in the 1425 which were tested before this month and didn't appear before.)

The most interesting thing to me is that I used to have 25 predicted second cousins, none of who actually are, and most of which I can't even estimate a relationship. (If they had all answered my emails, it might have helped. ) Anyway, I now have 14 predicted second cousins, one of which is a new match from a test completed this month. The other 13 are retained matches from earlier.

The leaves 12 old matches. Of those 12, 2 are now predicted 4th cousins, 7 are 4th-remote, and 3 are 5th-remote. I didn't check every case, but in those I checked, both the longest block and shared cM were lower than before.

[MarP]

Match loss hitting Finns'

This change hits Finns especially badly, many lost 30-50 % of distant cousins. Many Finns know family history back to the 1500's and 1600's, thanks to excellent church registers and digitalized archives. We have excellent cooperation within our Finland DNA project community, so we have been able to solve many connections by seeing patterns in common segments, by grouping distant cousins etc. Now these meaningful segments have disappeared, even quite close relatives have lost considerable segments and known distant cousins have disappeared.

Please give us back our common segments and lower tresholds so that we can continue our fruitful genetic genealogy & trad. genealogy cooperation.

[Naughtius]

Hey I would love to find online phonebooks of individual Irish villages! Can you suggest where to find these online??

I would suggest finding a name of a local news paper in the area you are researching and write a letter there first and also maybe send a letter to the local parish church.

[DSproed]

Gedmatch and "adding" people known to match

I appreciate the flexibility of uploading the data to Gedmatch for additional opportunity to control the matches. I have several "distant" matches. What I'd like is to be able to "add them back" to the matches list when we know how they match on paper.

I had some previous to the upgrade that I found thru Gedmatch, granted distant beyond 5th generation but still I could identify them.

What I don't understand is why not everyone uploads their data to Gedmatch.

Denise

[MoberlyDrake]

I wouldn't want the cuttoff point to be lowered too much!

On the one hand, it's frustrating to have only 124 matches here.

On the other hand, I have 2300 matches at Ancestry, with a warning that there is a 50% or less chance that I'm related to any one of them!!! I'd rather have fewer matches with a greater likelihood of a relationship.

Curiously, among my last 20 matches an old Affy match has appeared together with 9 matches who weren't there before, and the 9 are all in common with the old Affy match, plus there are 2 matches I've had for quite a while who are in common with him. I compared surname list for those of the 12 in common matches who had them and didn't notice anything interesting.

[T E Peterman]

The odd thing about all of this is that we had been hearing for months from both Family Tree DNA & other sources that Build 37 would help refine the quality of matches. For some reason, a lot of participants thought that this would mean more matches.

But "refining the quality" suggests that some of the Build 36 matches were not really good, just background noise, & through Build 37's refinement, some of the background noise would be eliminated.

So Build 37 goes in & everyone loses their highly speculative/ borderline noise matches, and everyone complains...

The only complaints that I think are noteworthy are the cases where people are no longer matched to close relatives (ie, siblings & such). The message that all of us have from Bennett Greenspan indicates that these are being addressed & re-run & that, some time this week, these problems should be taken care of.

Timothy Peterman

[ToddH]

I prefer the Ancestry method which allows you to discover a connection even to "low confidence" matches. When I find a common ancestor with anyone, however remote the genetic connection seems to be, that's one more cousin that I can know and compare notes with. If I find no connection, it's no skin off my teeth.

I don't understand what we've got to lose by "lowering the bar" for matches. The more the merrier, it increases the likelihood of finding connections to relatives. Isn't that what we paid for? I don't understand the "Keep them out!" mentality toward people who may not be completely authentic matches. I think FTDNA is trying too hard to make sure matches are "real." I want to be given the chance to connect with people who fall into the "gray area."